Abstract

A number of methods available for comparing two samples with censored data is evaluated through a simulation-based exercise. The (Geweke-Hajivassilion-Kenne (GHK) simulator is used here. All the methods discussed here can handle the case when there are multiple detection limits. Under the conditions considered in the simulation, the Mann-Whitney/ Wilcoxon method is best in maintain the Type 1 error rate, while still providing sufficient power.